BGP EVPN VTEP learning and nexthop behavior with EBGP sessions

[Girish Nagaraj]

Hi,

The VTEP learns about other VTEPS, is it only via the originating router IP in the T3 route prefix?

How the T2 route nexthop is handled in the clos topology with E-BGP sessions between leaf & spines, is it modified at every hop ?

Thanks

[Pavel Tarakanov]

>How the T2 route nexthop is handled in the clos topology with E-BGP sessions between leaf & spines, is it modified at every hop ?

There are several models of using EBGP as underlay, but all of them preserve next-hop (on spine in spine-leaf topology) so VXLAN tunnel build between VTEPs.

[Girish Nagaraj]

My question is regarding BGP EVPN overlay sessions between spine & leaf nodes being E-BGP sessions, in this case the MAC-IP route received by the spine node, does it set nexthop to self before forwarding this update message ?

[M02@rt37 · Accepted solution]

Hello @Girish Nagaraj 

BGP next-hop is not changed for EVPN routes. When spine receive a Type2 MAC/IP route from a leaf, it re-advertises the route to other leafs with the originating VTEP’s loopback still as the next-hop, not its own.

So, all remote VTEP can tunnel directly to the advertising VTEP since the underlay provides IP reachability to every loopback.  Unlike "normal" eBGP behavior where nexthop self is set, EVPN require preservation of the VTEP lopback as next-hop...

[Pavel Tarakanov]

https://www.cisco.com/c/en/us/td/docs/dcn/nx-os/nexus9000/103x/configuration/vxlan/cisco-nexus-9000-series-nx-os-vxlan-configuration-guide-release-103x/m_configuring_vxlan_bgp_evpn.html#concept_cjq_y3n_yfb

Please find above page as example of configuring EBGP.

[Giuseppe Larosa]

Hello @Pavel Tarakanov ,

the next hop unchanged is achieved using the following route-map applied to neighbors in AF l2vpn

route-map NEXT-HOP-UNCH permit 10
  set ip next-hop unchanged

So it is not an automatic setting in AF l2vpn evpn ?

Best Regards

Giuseppe

[Pavel Tarakanov]

No, AF l2vpn follows standard BGP rules and change next-hop for eBGP. It used, for example, in multi-site setups on BGWs.

But in case of eBGP in single site we need to change this default behavior to build direct VXLAN tunnels between VTEPs.